Development And Validation Of A Clinical Breast Cancer Tool For Accurate Prediction Of Recurrence

This article discusses the use of machine learning methods to predict high-risk Oncotype DX (ODX) status based on readily available clinicopathologic features. These models have been developed using patient data from the National Cancer Database (NCDB) and have been validated in various contexts. However, until recently, the NCDB did not report quantitative histologic parameters, limiting the accuracy of these models. The article presents the results of training machine learning models in the NCDB incorporating these quantitative histologic variables and validating them in a large and diverse patient cohort from the University of Chicago Medical Center (UCMC).